[ 
https://issues.apache.org/jira/browse/MAPREDUCE-5356?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13695550#comment-13695550
 ] 

Ashwin Shankar commented on MAPREDUCE-5356:
-------------------------------------------

Thanks for your comment Devaraj.
So we then have two options here :
1. We don't include LOG_AGGREGATION_ENABLED property as part of the refresh.
Currently in trunk if you look at 
AggregatedLogDeletionService#serviceStart(),we schedule the logDeletionTask 
only if
LOG_AGGREGATION_ENABLED(LAE) is set to true. Then I would need to the same 
check at the start of my refresh code ie dont refresh 'log retention','check 
interval' prop AND dont start the deletion task if LAE property is false. 
But this means - '-refreshLogRetentionSettings' depends on LAE and would be 
useless until we start history server with LAE set to true, AND
we need to bounce the history server and NM if we want to change LAE.

2. Second option, is to make two rpc calls -one to HS and one to NM to do the 
refresh on LAE. But as far as I looked, I don't see an admin interface on NM to 
do that.
We could probably create another JIRA to create an RPC server on NM and have a 
separate command something like '-refreshLogAggregationEnable' .

What are your thoughts ?
I'll post a new patch based on my point-1 if thats agreeable.

                
> Refresh Log aggregation 'retention period' and 'check interval' 
> ----------------------------------------------------------------
>
>                 Key: MAPREDUCE-5356
>                 URL: https://issues.apache.org/jira/browse/MAPREDUCE-5356
>             Project: Hadoop Map/Reduce
>          Issue Type: Sub-task
>          Components: jobhistoryserver
>    Affects Versions: 2.1.0-beta
>            Reporter: Ashwin Shankar
>            Assignee: Ashwin Shankar
>              Labels: features
>         Attachments: MAPREDUCE-5266-2.txt, MAPREDUCE-5266-3.txt
>
>
> We want to be able to refresh log aggregation retention time
> and 'check interval' time on the fly by changing configs so that we dont have 
> to bounce history server.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

Reply via email to